Shortly after he served on a jury in March, Gregg Crumley developed a sore throat and congestion. The retired molecular biologist took a rapid test on a Saturday and saw a dark, thick line materialize - "wildly positive" for the coronavirus.
Crumley, 71, contacted his doctor two days later. By the afternoon, friends had dropped off a course of Paxlovid, a five-day regimen of antiviral pills that aims to keep people from becoming seriously ill.
The day he took his last dose, his symptoms were abating. He tested each of the next three days: all negative.
Then, in the middle of a community Zoom meeting, he started feeling sick again. Crumley, who is vaccinated and boosted, thought it might be residual effects of his immune response to the virus. But the chills were more prolonged and unpleasant. He tested. Positive. Again.
Video: What we know about long covid, two years into the pandemic
Crumley, like other patients who have experienced relapses after taking Paxlovid, is puzzled - and concerned. On Twitter, physicians and patients alike are engaged in a real-time group brainstorm about what might be happening, with scant evidence to work with.
It is the latest twist - and newest riddle - in the pandemic, a reminder that two years in, the world is still on a learning curve with the coronavirus.
Infectious-disease experts agree that this phenomenon of the virus rebounding after some patients take the drug appears to be real but rare. Exactly how often it occurs, why it happens and what - if anything - to do about it remain matters of debate.
What's clear is that patients should be warned it is possible so they don't panic - and so that they know to test again if they start feeling ill. More data is needed to understand what is going on. Paxlovid, made by the drug giant Pfizer, remains a useful drug, even though it has sparked a new mystery.
"I'm not negative on Paxlovid," said Crumley, who lives in Philadelphia and whose last positive test was a week after his second wave of illness began. "I don't know whether it's just stopping [viral] replication for that five-day period of time, and it comes back."
One of the top worries accompanying antiviral drugs is the threat of resistance, when the virus evolves to evade the treatment. A Food and Drug Administration analysis of Pfizer's clinical trial of the drug showed the virus rebounded in several subjects about 10 to 14 days after their initial symptoms but found no reason and no clear evidence that their infections were resistant to the treatment.
Michael E. Charness, chief of staff at the VA Boston Healthcare System, published a detailed case study of one 71-year-old patient who had a relapse. The man, who was vaccinated and boosted, received Paxlovid and quickly felt better. When he developed cold symptoms a week after his case of covid had resolved, researchers sequenced the virus's genetic code and found it was the same virus surging back. That ruled out a reinfection, the emergence of a variant or the virus becoming resistant.
Charness would like to see more data and other questions answered. Should antivirals be given longer, to assure the virus is cleared? Should people be treated a second time? What are the implications for people returning to their normal lives?
"If you have a resurgence of viral load, and that happens on day 10, when CDC says you're back to work, no mask, what are you supposed to do about isolation? Is that a moment when you're contagious again?" Charness said. "The person we studied, we advised to isolate until their viral load was gone the second time."
Pfizer is collecting data, in clinical trials and in real-world monitoring of the drug's use. The company's trial data indicates there is a late uptick in viral load in some people who take the drug, but the rates appear to be the same among study participants given a placebo, according to a person familiar with the data who spoke on the condition of anonymity to speak freely.
Those findings suggest that Paxlovid isn't the reason people are relapsing, because that's happening in untreated people, too.
If that turns out to be true, it raises the concern that some people - whether they have taken the drug or not - could be infectious long after they think they are in the clear, and after guidelines suggest they can stop taking precautions.
"We have not seen any resistance to Paxlovid, and remain very confident in its clinical effectiveness," Pfizer spokesman Kit Longley wrote in an email.
The limited evidence leaves most physicians favoring the idea that Paxlovid knocks the virus down but doesn't knock it out completely. It's possible that by holding the virus in check, the immune response doesn't fully ramp up, because it doesn't see enough virus. Once the treatment ends, the virus can start multiplying again in some people.
Philip Bretsky, a primary care doctor in Santa Monica, Calif., said he has encountered two cases among patients, both of whom were vaccinated and boosted at least once.
A double-boosted 72-year-old who had chronic medical conditions that raised his risk for severe illness started to feel unwell at the end of March. He tested positive and began a course of Paxlovid. He felt better and tested negative. Then, 12 days later, he started feeling crummy again - and tested positive.
Reinfection seemed improbable, and Bretsky thought resistance was unlikely with a five-day course of treatment.
In well-vaccinated people, being reinfected so quickly would be "like getting struck by lightning or winning the lottery," Bretsky said. "I don't think this is reinfection. I think this is recrudescence of the original infection."
Experts don't know how common this phenomenon is. Many people may not test if they get sick again after their initial infection has receded, making it hard to track.
That almost happened to Holly Teliska, 42, of San Francisco. Teliska got sick shortly after returning home from a trip to New York. She has a risk factor for severe illness and got access to Paxlovid right away. When she finished her treatment course, she took a home PCR test that was negative and felt much better, though remained fatigued.
Four days later, she came down with a runny nose and cough. She assumed she had caught her daughter's cold and powered through. Five days later, with plans to visit an immunocompromised friend, she took a test.
Teliska almost felt silly testing herself. She had been vaccinated and boosted, then infected.
"We've been saying I'm her safest friend now, now that I've had covid, so for three months, I can go spend time with her pretty safely," Teliska said. "That really threw that narrative out the window. . . . This entire experience has been a real reminder there is still so much to learn."
Paxlovid is new. It only began to be used in December, so reports people share on social media of resurgent illness may be the tip of the iceberg - or might simply reflect the eagerness to learn more about a rare, intriguing outcome.
If such cases turn out to be exceedingly rare, then these case reports may be a sporadic curiosity - something to warn patients could happen. If more common, it could lead to tweaks in treatment regimens.
The mounting anecdotes are compelling to many physicians, but it's also possible the virus might rarely rebound. Yonatan Grad, an associate professor of immunology and infectious diseases at the Harvard T.H. Chan School of Public Health, has studied the viral loads of NBA players and staff during the course of an infection. That data, he said, shows that viral loads can bounce around.
What's "exceptionally uncommon," Grad said, is for the viral load to plunge for a few days to a level that suggests they are negative and then go up again.
Paul Sax, an infectious-diseases specialist at Brigham and Women's Hospital in Boston, recently shared the story of a patient who became infected and then relapsed after taking Paxlovid. He has heard from lots of colleagues with similar stories. But the anecdotes raise more questions than they answer.
Even if the virus has not been shown to develop resistance to the treatment during a resurgence, that's doesn't mean it won't happen, he points out. Does the treatment knock the virus down so successfully that people aren't generating a robust immune response? That could have implications for understanding whether being infected acts as a potent booster.
The phenomenon is so new that many doctors aren't aware of it. Jennifer Charness, a 31-year-old nurse who lives in Brookline, Mass., had the benefit of knowing about her father's work at the Boston VA.
Charness started sneezing in early April and got a blaringly positive coronavirus test. She has a history of asthma and was prescribed Paxlovid. As she took the drug, she saw her positive test line grow fainter and her symptoms resolve. She swabbed to make sure she was negative before going back to work, as a precaution. Then, two days later, she felt the symptoms come back and tested positive - again.
"I'm so frustrated," Charness said. "I don't think I'm going to get very sick. It's the concern of what does this mean for my viral load, and how contagious am I? And when will I not be contagious? I'm stuck back in my home again."
Charness's primary concern is that she doesn't pose a risk to anyone else. She consulted a doctor via telemedicine Friday. The practice hadn't heard of any cases like hers and decided to treat it as a reinfection and reset the isolation clock.
"I'm Day 1," she said. "Or am I Day 13?"